MutableMapping implementations for common cloud storage providers

For now Azure Blob Storage, Google Cloud Storage, and AWS S3 are implemented. Contributions of new providers are welcome.

with pip:

pip install cloud-mappings

By default, cloud-mappings doesn't install any of the required storage providers dependencies. If you would like to install them alongside cloud-mappings you may run any combination of:

pip install cloud-mappings[azure,gcp,aws]

from cloudmappings import AzureBlobMapping

cm = AzureBlobMapping.with_pickle(
    account_url="AZURE_BLOB_STORAGE_URL",
    container_name="CONTAINER_NAME",
    credential=AZURE_CREDENTIAL_OBJECT,
)

from cloudmappings import GoogleCloudStorageMapping

cm = GoogleCloudStorageMapping.with_pickle(
    project="GCP_PROJECT",
    credentials=GCP_CREDENTIALS_OBJECT,
    bucket_name="BUCKET_NAME",
)

from cloudmappings import AWSS3Mapping

cm = AWSS3Mapping.with_pickle(
    bucket_name="AWS_BUCKET_NAME",
    silence_warning=False,
)

Note that AWS S3 does not support server-side atomic requests, so it is not recommended for concurrent use. A warning is printed out by default but may be silenced by passing silence_warning=True.

Use it just like a standard dict()!

cm["key"] = 1000
cm["key"] # returns 1000
del cm["key"]
"key" in cm # returns false

Each cloud-mapping keeps an internal dict of etags which it uses to ensure it is only reading/overwriting/deleting data it expects to. If the value in storage is not what the cloud-mapping expects, a cloudmappings.errors.KeySyncError() will be thrown. If you know what you are doing and want your operation to go through anyway, you will need to sync your cloud-mapping with the cloud by calling either .sync_with_cloud() to sync all keys or .sync_with_cloud(key) to sync a specific key. By default .sync_with_cloud() is called on instantiation of a cloud-mapping if the underlying provider storage already exists. You may skip this initial sync by passing an additional sync_initially=False parameter when you instantiate your cloud-mapping.

If you don't call .with_pickle() and instead pass your providers configuration directly to the CloudMapping class, you will get a "raw" cloud-mapping which accepts only byte-likes as values. Along with the .with_pickle() serialisation utility, .with_json() and .with_json_zlib() also exist.

You may build your own serialisation either using zict; or by calling .with_buffers([dumps_1, dumps_2, ..., dumps_N], [loads_1, loads_2, ..., loads_N]), where dumps and loads are the ordered functions to serialise and parse your data respectively.